breathy voice is phased after modal voice as a means to insure optimal recoverability of tonal contrasts. She terms such a phasing, the “Solomonic effect” 553.
Regarding Zapotec creakiness, Herrera challenges Silverman’s claim that non-modal voice features cannot co-occur with tone in Otomanguean languages Silverman 1995: 92, Herrera 2001: 558. This is
demonstrated with Zapotec where she concludes that “the aperture of the creaky vowel is reduced, but still enough to allow for laryngealization, [so that] tone can continue also” 558.
4.3.1.2 Autosegment
The second approach to laryngeally ambiguous vowels is represented in analyses such as those by Hollenbach 1984, McCaulay and Salmons 1995 and Meechan 1990 where the authors claim
laryngealization to be a floating feature based on Autosegmental phonology. Hollenbach, for example finds evidence for tonal alternations with post-vocalic h ʔ, while laryngeal onsets clearly pattern as
consonantal segments 30-31. Based on this evidence, Hollenbach posits both laryngeal autosegments those that interact with tone on a laryngealtonal tier and laryngeal segments h ʔ.
McCaulay and Salmons 1995 observe that glottal stop is contrastive only on Mixtec roots. Since Mixtec may have roots such as CVʔ and CVʔV they argue that a floating glottal stop feature attached to
the left vowel of a root will simplify earlier accounts, which posit a contrast between a glottalized and plain vowel 49. They conclude that where a vocalic glottal stop would double the vowel inventory
58, their analysis allows for a more economical and hence, preferable approach to Mixtec glottalization.
4.3.1.3 Root-plus-formative
With regard to Northern Pame, only Avelino 1997 broaches the subject of laryngeally ambiguous vowels. Regarding such forms as those in figures 4-4 and 4-5, Avelino maintains that these are best
understood as a “formative” on the root to form a stem 109. However, he gives no definition as to the meaning of this formative, if indeed it has any morphological meaning at all.
4.3.2 Critique of former approaches
Laryngeally ambiguous vowels in Northern Pame appear quite similar to those outlined in the unit and Autosegmental approach for Trique, Zapotec, Amuzgo and Mixtec, yet there are a number of unanswered
questions to these previous analyses. With regard to the unit approach proposed by Silverman 1995, the postulation of underlying breathy and creaky vowels depends on one crucial phenomenon, the
presence of tone. However, data exist from non-Otomanguean as well as Otomanguean languages that prove that tone can be produced simultaneously with non-modal creaky voice, as Silverman as well as
Herrera both mention. Regarding the reason as to why certain Otomanguean languages do not allow simultaneous tone and non-modal voice, Silverman and Herrera both claim that laryngeal and pitch
gestures are both produced in lesser degree and hence, they are able to co-occur Silverman 1997b: 247, Herrera 2001: 558. This argumentation appears to be unconvincing and in fact circular tone and voice
quality are incompatible thus triggering laryngeal sequencing, except in languages where they are compatible.
28
That is not to say that laryngeally ambiguous vowels are not units in these languages, but only that the postulation of tone as the key cause for laryngeal sequencing is yet to be definitively
proven. As the data of Northern Pame demonstrate below, this present research maintains that syllable phonotactics must also be considered.
The Autosegmental approach, while attempting to simplify the syllable complexity of the laryngeal
28
In fact, other Mexican language exhibit laryngeal vowel sequencing that are not tonal. For example, Isthmus Mixe Herrera, p.c. has this characteristic with no contrastive tone.
root in Mixtec and Trique, does so by adding a new [auto]segment to the inventory. This approach has merit in their data set, since Mixtec apparently has no true glottal stop segments and laryngeal segments
in Trique are limited to onsets.
29
However, Northern Pame laryngeals are clearly segmental in the phonology, appear outside the root, and do not alternate with tone. Thus, an Autosegmental approach
does not have direct application to the data facts in any way that seems directly useful. Finally, the root-plus-formative approach maintained by Avelino 1997 clearly sees these vowels
and laryngeals as sequences of separate segments. He claims that the second vowel is a reduplicant or copy of the first Avelino 1997: 109 in much of the same spirit as other Pame linguists Soustelle 1937,
Gibson 1956, Manrique 1967. However, while this research agrees with interpreting these data as vowellaryngeal sequences, the data in the following analysis suggest that the first vowel rather than the
second is the copy vowel.
4.3.3 Reanalysis of laryngeally ambiguous vowels